Vision Guided Food Assembly by Robot Teaching from Target Composition


Automation in the food-serving industry has challenging requirements. The article deals with the application of vision-guided food assembly of lunch box containers by robot teaching from a demonstration of a given target composition for the container. Vision guidance is used for parsing the target composition, bin picking with instance segmentation, and computing pose information for similarity-based assembly for food serving with tracking of containers on a moving conveying system. A control system based on learning from demonstration technique is employed in the object frame for similar pose object dropoff of the grasped item for assembly. Integration of vision-guided robot control into high-speed automated food assembly can be highly productive. A pilot experiment scenario for automated food assembly demonstrates the functionality of the vision-guided food-serving system.